论文信息 - Modeling a Multi-Agent Environment Combining Influence Diagrams

Modeling a Multi-Agent Environment Combining Influence Diagrams

Abstract In this paper we show how inﬂuence diagrams (IDs) can be combined for modeling a multi-agent environment. Inﬂuence diagrams are a compact representation of a joint probability dis-tribution where actions and utilities are explicitly modeled.Combining diﬀerent IDs which represent individual agents into a global coherent networkrepresenting the entire multi-agent system is a diﬃcult problem. For this purpose we introducetwo approaches and compare them. 1 IntroductionIn this paper we address the problem of modeling the environment and other agents inmulti-agent systems (MAS). Existing formalisms such as the Markov game model [1, 2]suﬀer from combinatorial explosion, since they learn values for combinations of actions ofall the agents. We suggest to use inﬂuence diagrams [3] (also known as decision networks)to avoid this problem of tractability.We will discuss the problem of modeling the environment and other agents acting in theenvironment in the context of Markov Games [1, 2]. The Markov game model is deﬁnedby a set of states S, and a collection of action sets A

[1] Peter Norvig,et al. Artificial Intelligence: A Modern Approach , 1995 .

[2] Tom Lenaerts,et al. Niching and evolutionary transitions in multi-agent systems , 2001 .

[3] David Maxwell Chickering,et al. Dependency Networks for Inference, Collaborative Filtering, and Data Visualization , 2000, J. Mach. Learn. Res..

[4] Richard S. Sutton,et al. Reinforcement Learning: An Introduction , 1998, IEEE Trans. Neural Networks.

[5] D. Heckerman,et al. Dependency networks for inference , 2000 .

[6] Michael P. Wellman,et al. Multiagent Reinforcement Learning in Stochastic Games , 1999, ICML 1999.

[7] Frank Jensen,et al. From Influence Diagrams to junction Trees , 1994, UAI.

[8] Anders L. Madsen,et al. LAZY Propagation: A Junction Tree Inference Algorithm Based on Lazy Evaluation , 1999, Artif. Intell..

[9] K. Tuyls,et al. Niching and Evolutionary Transitions in MAS , 2007 .

[10] Nevin Lianwen Zhang,et al. Probabilistic Inference in Influence Diagrams , 1998, Comput. Intell..

[11] Michael L. Littman,et al. Markov Games as a Framework for Multi-Agent Reinforcement Learning , 1994, ICML.

[12] Judea Pearl,et al. Probabilistic reasoning in intelligent systems - networks of plausible inference , 1991, Morgan Kaufmann series in representation and reasoning.